Economist Tyler Cowen argues that the "least bad" way to regulate artificial intelligence is to trust the major AI labs to act as their own primary regulators, through a private oversight body authorized and supervised by Washington.

The argument appears in his latest Free Press column, excerpted on the economics blog Marginal Revolution on August 27, 2026. "The key is to create some basic safeguards, but without stifling broader AI progress," Cowen writes. To do so, he contends, "we must defy the conventional wisdom about public oversight and instead trust the AI labs to be their own primary regulators."

A FINRA-style regulator for AI

The proposal starts by defining a private, not-for-profit body for AI regulation. An ideal body would draw features from FINRA (the Financial Industry Regulatory Authority) — a consortium of financial firms that examines the trade practices of each member and makes recommendations, helping the federal Securities and Exchange Commission with oversight and regulation.

The AI version would include the major labs and would be authorized and overseen by Washington, perhaps through the now-fledgling Center for AI Standards and Innovation. In Cowen's telling, that would give policymakers a way to put guardrails around advanced models without requiring government agencies to build all of the relevant technical expertise from scratch.

Periodic audits and a liability exemption

This body would periodically audit major AI companies and their models, judging their conduct and safety. In the short run at least, much of that work would focus on cybersecurity — in particular, whether the new models created more cyber risk than they help to solve.

A company that passed the audit would be exempted from standard liability law, provided it had shown basic, reasonable care rather than extreme or deliberate negligence. In Cowen's view, that would free the AI labs from the fear that courts might derail their business by granting huge awards to plaintiffs for ill-defined harms that could not reasonably have been prevented — and it would give the labs a strong incentive to meet the safety standards of this body.

Would an industry body judge fairly?

It is reasonable to wonder, Cowen notes, whether a body composed of industry players would issue fair and equitable judgments of safety. "Maybe not. Yet there are many upsides and no better alternative."

One upside: each company knows that a dangerous model from another company could cause a harmful incident and damage the prospects of the entire industry. Cowen points to the Three Mile Island meltdown in 1979, which contributed significantly to the mothballing of the entire U.S. nuclear industry. Few people can name the company behind the malfunctioning plant — Metropolitan Edison — because the reputational penalty attached to the industry as a whole.

Another incentive for safety is that the top companies do not want too much competition from lower-price, lower-quality upstarts. That, too, will induce those companies to support fairly tough standards, "perhaps excessively tough in some cases."

The broader implication of Cowen's argument is that AI governance may end up depending less on a traditional rulebook than on how industry participants, regulators, and courts divide responsibility as the technology keeps advancing. For now, he argues, that makes technical expertise inside the labs especially important to any oversight regime that aims to be enforceable rather than symbolic.

"Still, we are choosing from imperfect alternatives," Cowen concludes. "The concrete truth, whether we like it or not, is that there is far more expertise within the companies for judging AI safety than we can expect to find in the federal government anytime soon."

Acknowledgments

Cowen writes that he is indebted to some ideas from Dean Ball, noting that Ball's proposal is somewhat different, and he points readers to comments from Brendan McCord.